Hi, Oleksandr,
On Oct 19, Oleksandr Byelkin wrote:
> On Wed, Oct 19, 2022 at 5:09 PM Sergei Golubchik <serg@mariadb.org> wrote:
> > On Oct 19, Oleksandr Byelkin wrote:
> > >
> > > if (share->view_def->parse((uchar *) &share->tabledef_version, NULL,
> > > view_timestamp_parameters, 1,
> > > &file_parser_dummy_hook))
> > > + {
> > > + // safety if the definition file is brocken
> > > + share->tabledef_version.length= 0;
> >
> > when can view_def->parse() fail? _Only_ if the definition is invalid?
> > Can it succeed for one TABLE and fail for another, both of the same
> > TABLE_SHARE?
>
> Also if there is no memory.
>
> If for the other table there will be more memory or file has changed
> it can succeed.
My point is that if it is possible that two different threads would get
different results from share->view_def->parse(), then you cannot update
the shared share->tabledef_version.length without a lock. And you cannot
read it without a lock.
So if it is possible that two different threads would get different
results from share->view_def->parse(), then it's best not to update
share->tabledef_version.length here and only do it in open_table_def().
if we got an error as I understood it will nead to an error of opening the table, so the query will fail.
The other thread may have memory but it is true for any EOM situation and there are a lot of places where we allocate memory, so I do not see how this will differ.
The same with changing frm file with a working server, I do not think that our server will survive if something out of the server will change frms.
Regards,
Sergei
VP of MariaDB Server Engineering
and security@mariadb.org